Update Domain Knowledge

Incrementally update your SciAgent domain configuration. Use this after
the initial /configure-domain setup to add new packages, refine
workflows, or extend template content without starting from scratch.

When to Use

• You want to add a new package to your domain configuration.
• Your analysis workflows have changed and need updating.
• You discovered a new tool or library relevant to your research.
• You want to refine value ranges, parameters, or guardrails.
• You have new example data that reveals additional file formats or
  column patterns.

Procedure

Step 1 — Understand the Change

Ask the user what they want to update. Common scenarios:

• "Add package X" — Discover, document, and integrate a new
  package into the template files.
• "Update workflows" — Modify or add workflow steps in
  workflows.md.
• "Change parameters" — Update analysis parameters, value ranges,
  or precision tables in operations.md.
• "Add a custom skill" — Create a new skill entry in skills.md.
• "Full refresh" — Re-audit all templates and fill any remaining
  placeholders (suggest /configure-domain instead if most content
  is missing).

Step 2 — Audit Current State

Scan the relevant template files to understand what's already configured:

1. Search for files matching *.instructions.md, operations.md,
   workflows.md, tools.md, library_api.md, skills.md.
2. Identify which sections have domain content vs. unfilled
   <!-- REPLACE: ... --> placeholders.
3. Note any domain-specific content that might conflict with the
   proposed changes.
4. Brief the user: "Here's your current configuration in the affected
   files. I'll update X and Y."

Step 3 — Discover (If Adding Packages)

If the user wants to add new packages:

1. Fetch PyPI metadata: https://pypi.org/pypi/{name}/json
2. If available, fetch the GitHub README for capabilities overview.
3. Present the package info and ask for confirmation.
4. If the user doesn't know the exact package name, use broader
   searches based on their description.

Step 4 — Propose Edits

Before making changes, show the user exactly what will be modified:

• Which files will be edited
• What content will be added or changed
• Whether any existing content will be affected

Ask for confirmation before proceeding.

Step 5 — Apply Updates

Use editFiles to make the confirmed changes:

• Adding a package: Update library_api.md with API reference,
  tools.md with tool documentation, and operations.md if the
  package introduces new analysis parameters or workflows.
• Updating workflows: Edit the relevant section in workflows.md.
• Changing parameters: Edit the parameters table in operations.md.
• Adding skills: Add a new skill section to skills.md and
  optionally create a SKILL.md file in .github/skills/.

Step 6 — Verify

1. Confirm the edit was applied correctly by reading the modified file.
2. Summarize what was changed.
3. Note any follow-up actions (e.g. "You may want to run
   /docs-ingestor on the new package for a deeper API reference").

Re-Run Safety

• Preserve existing content — Only modify the specific sections
  relevant to the update. Do not touch unrelated template content.
• Ask before overwriting — If the target section already has
  different domain content, ask the user whether to replace, merge,
  or skip.
• Append by default — When adding new items (packages, workflows,
  skills), append to existing lists rather than replacing them.

What This Skill Does NOT Do

• Does not execute Python code or run analysis
• Does not install packages (suggest pip install commands instead)
• Does not require sciagent[wizard]
• Does not replace the full /configure-domain flow for first-time
  setup — if most template content is missing, suggest running
  /configure-domain first

Domain Customization